Drug Discovery

Accelerating compound screening and binding estimation via advanced spatial algorithms and graph representations.

Binding Affinity Prediction

Estimating the binding strength of small-molecule ligands to target protein receptors using molecular graphs and docking simulators.

• Graph Neural Networks (GNNs): Representing atoms as nodes and bonds as edges, utilizing message passing to compute chemical properties.
• Cross-Attention Mapping: Implementing cross-attention mechanisms between target protein residues and ligand atoms to resolve interaction interfaces.

Active Learning Pipelines

Creating workflows to iteratively query simulation software (e.g., AutoDock Vina) for labels, and training models to screen databases of millions of compounds.